You may have run into the terms generative AI or large language models, whether in your work space or social space, but there is another concept that is being talked about in AI circles that AI practicioners are really excited about called Agentic AI.
So what exactly is an AI agent?
Agentic AI is a combination of software that uses artificial intelligence to pursue specified goals. It accomplishes this by decomposing the goal into actionable tasks, monitoring its own progress, and engaging with other digital resources (search, internal knowledge bases, specific tools), and even other agents!
One of the key differences between Agentic AI and simply asking for a response from a conversational interface like ChatGPT or Bard, is that the agent can take a higher level goal, break it down into sub-tasks and execute it on, with you having to give it a bunch of discrete instructions on how to do each task.
Here is an example of what this might look like:
User Objective:
Find me the best candidate for this role out of all the interviews conducted Agent Determines and Executes Each of These Sub-Tasks:
Review Internal Information
Understand the job description to figure out key criteria
Review any comments by recruiters and hiring managers about this role.
Review interviews from past similar roles, where candidates were successfully placed and analyze any patterns that could be applied to future candidates.
Review Current Candidate Information
Review each candidate resume
Performs a web search on the candidate
Analyze each call
Determine the best candidate
Create a summary for each candidate
Score their responses against a matrix
Rank the top 5 candidates
Create an executive summary of the top 3 candidates weighing the pro’s and con’s.
Hopefully, by now you get the idea. Agents can go even granular in terms of sub-tasks than shown above, but the major concept is that you just have to set a goal, and the agent will create the sub-tasks and get it done.
Comments